Exhibit KE-1A
Technical Discussion of the Postal Service’s
Understatement QBRM cost savings
Mr. Miller’s corrected QBRM unit cost savings --1.6 cents -- is less than one-third of the 5.0-cent cost savings that I derive. There are three separate flaws that lead to such disparate results. First, Mr. Miller omits from his analysis all other additional operations after the outgoing primary. His explanation for this omission is that the cost savings should only reflect the Postal Service’s cost of applying a barcode to a handwritten addressed envelope. See response to KE/USPS-T22-28A. However, Mr. Miller’s position disregards two important facts that have a direct bearing on the proper derivation of cost savings: (1) the Postal Service cannot possibly barcode 100% of HAND letters, and (2) pursuant to USPS requirements, 100% of QBRM letters must be prebarcoded, and the barcodes and other attributes affecting processing costs must be pre-approved. Therefore, Mr. Miller’s analysis omits QBRM savings that accrue downstream since, after the outgoing primary, the proportion sorted by automation will be greater for QBRM letters than for HAND letters.
Second, Mr. Miller’s model understates the unit processing cost for letters that are processed through the RBCS operation and, at the same time, overstates the unit processing cost for letters that bypass the RBCS, such as QBRM. Both of these points are clearly illustrated by the comparison of model-derived and CRA-derived unit costs shown in Table 4 of Exhibit KE-T-1. USPS witness Miller should have recognized these inconsistent results exhibited by his model-derived unit costs.
Finally, Mr. Miller compounds the inconsistent results exhibited by his models by inappropriately applying the same BMM CRA proportional adjustment factor for both the HAND and QBRM models. As discussed above, when the entry point for a rate category is the RBCS, the model will understate costs. The BMM CRA adjustment factor corrects this problem by raising the model-derived unit cost. But applying this same BMM CRA adjustment factor to QBRM, which bypasses the RBCS, only makes the problem of overstating QBRM costs worse. Therefore, in order to accurately tie the model-derived unit cost to the CRA, Mr. Miller should have applied the Automation CRA adjustment factor to the QBRM model-derived unit cost.
In his Direct Testimony, Mr. Miller did not provide any reason to support his application of the BMM CRA adjustment factor for both HAND and QBRM. Belatedly, in response to a KeySpan interrogatory, he noted that QBRM and HAND letters “are subsets of the First-Class single-piece mail stream.” See response to Interrogatory KE/USPS-T22-8C. Mr. Miller’s observation, in this instance, is not very useful. While the processing mail flows of HAND and BMM letters are very similar, the processing mail flows for QBRM and BMM letters are very different. QBRM is unique within the First-Class single piece mailstream because it is prebarcoded, machinable and possesses a complete and accurate address by definition. The fact that QBRM is mailed as single piece and makes up a tiny part of that subclass is simply not important. As discussed above, the most important cost determinant for single piece letters, as presented by Mr. Miller’s model, is whether or not the letters are processed in the RBCS operation. Non-automation presorted letters and BMM letters require RBCS processing. Therefore, it is inappropriate to apply either the non-automation[1] or the BMM CRA adjustment factor to reconcile the model-derived unit cost of QBRM that does not require RBCS processing.
Finally, there was good reason for Mr. Miller’s decision to omit processing operations after the outgoing primary after KeySpan pointed out various shortcomings in his original analysis. Had Mr. Miller followed the Commission’s methodology from Docket No. R2000-1, which included all operations up to and including the incoming secondary, Mr. Miller would have had even more problems that would have been difficult, if not impossible, to resolve.
Table 1 and Table 2 (below) are comparable to Tables 12 and 13 of Exhibit KE-T-1.[2] Table 1 shows a comparison of Mr. Miller’s CRA-reconciled workshare-related unit costs, had he not altered the model to stop the flow of mail after the outgoing primary.
Table 1
Comparison of Illustrative USPS-Derived Unit Costs
For QBRM to Other First-Class Rate Categories
(Cents)
(USPS Costs)
Hand Letters (Using BMM Adj Factor) / 8.33
QBRM Letters (Using BMM Adj Factor) / 7.49
Machinable Single Piece Letters / 6.45
Bulk Metered Mail (BMM) Letters / 6.45
Automation Mixed AADC Letters / 3.41
Automation AADC Letters / 2.68
Automation 3-Digit Presort Letters / 2.40
Source: KE-LR-J-4
As shown in Table 1, Mr. Miller’s reconciled workshare-related unit costs would have been anomalous. It is inconceivable that QBRM letters cost more to process than non-prebarcoded machinable single piece letters or BMM letters. Such a relationship simply does not make sense. Nor is it possible to reasonably explain the 4.08-cent difference (7.49 cents – 3.41 cents) between QBRM and Automation Mixed AADC letters.
The anomalous relationship shown in Table 1 is cured by applying the Automation CRA proportional adjustment factor to the QBRM model-derived unit cost, as shown in Table 2. The reconciled QBRM workshare-related unit cost (3.96 cents) is between the unit costs for Automation Mixed AADC (3.41 cents) and BMM letters (6.45 cents). This is where it should be.
Table 2
Comparison of Illustrative USPS-Derived Unit Costs
For QBRM to Other First-Class Rate Categories Using
KeySpan’s CRA Adjustment Factor
(Cents)
(USPS Costs)
Hand Letters (Using BMM Adj Factor) / 8.33
Machinable Single Piece Letters / 6.45
Bulk Metered Mail (BMM) Letters / 6.45
QBRM Letters (Using Auto Adj Factor) / 3.96
Automation Mixed AADC Letters / 3.41
Automation AADC Letters / 2.68
Automation 3-Digit Presort Letters / 2.40
Source: Library Reference KE-LR-J-4
Applying KeySpan’s recommended CRA adjustment factor to the Postal Service’s costs results in a sound, reasonable relationship.[3] Accordingly, I urge the Commission to reject Mr. Miller’s method for reconciling QBRM model unit costs to the CRA. It makes much more sense to apply the Automation letters CRA proportional adjustment factor to the QBRM model-derived unit cost.
Since the Postal Service’s QBRM model already overstates the derived unit cost, applying the BMM CRA proportional cost adjustment factor, as Mr. Miller does, only exacerbates the accuracy of the derived unit cost. His methodology mistakenly raises the already inflated model-derived unit cost by an additional 51%. (See USPS-LR-J-60, page 10)
The combination of these three errors explains why Mr. Miller understates QBRM savings by 3.4 cents per piece.
1
[1]In Docket No. R2000-1, the Commission applied the non-automation, presorted CRA adjustment factor to reconcile the QBRM model-derived unit cost. Based on the new model presented by the Postal Service in this case, this is no longer reasonable.
[2]The unit costs shown in Tables 12 and 13 reflect the Commission’s Docket No. R2000-1 methodology for measuring workshare-related unit costs. Those shown in Table 1 reflect the Postal Service’s proposed method for measuring workshare-related unit costs.
[3]Even if the Commission accepted the Postal Service’s cost attribution methodology and Mr. Miller’s proposal to eliminate two cost pools from the workshare cost savings analysis, the derived QBRM unit cost is 8.33 cents – 3.96 cents = 4.37 cents. This is still more than sufficient to justify the S&A’s proposed QBRM discount of 3 cents.